Scalable optical hypercube-based interconnection network for massively parallel computing

Two important parameters of a network for massively parallel computers are scalability and modularity. Scalability has two aspects: size and time (or generation). Size scalability refers to the property that the size of the network can be increased with nominal effect on the existing configuration. Also, the increase in size is expected to result in a linear increase in performance. Time scalability implies that the communication capabilities of a network should be large enough to support the evolution of processing elements through generations. A modular network enables the construction of a large network out of many smaller ones. The lack of these two important parameters has limited the use of certain types of interconnection networks in the area of massively parallel computers. We present a new modular optical interconnection network, called an optical multimesh hypercube (OMMH), which is both size and time scalable. The OMMH combines positive features of both the hypercube (small diameter, high connectivity, symmetry, simple routing, and fault tolerance) and the torus (constant node degree and size scalability) networks. Also presented is a three-dimensional optical implementation of the OMMH network. A basic building block of the OMMH network is a hypercube module that is constructed with free-space optics to provide compact and high-density localized hypercube connections. The OMMH network is then constructed by the connection of such basic building blocks with multiwavelength optical fibers to realize torus connections. The proposed implementation methodology is intended to exploit the advantages of both space-invariant free-space and multiwavelength fiber-based optical interconnect technologies. The analysis of the proposed implementation shows that such a network is optically feasible in terms of the physical size and the optical power budget.     

Electrophotonic computer networks with strictly nonblocking and self-routing functions

Three-stage optical interconnection networks for use in massively parallel processors are proposed. Wavelength-division- and space-division-multiplexing switches used in these networks are described, and free-space optics to assist in the construction of networks that are small and provide high throughput are discussed.     

Experimental free-space optical network for massively parallel computers

A free-space optical interconnection scheme is described for massively parallel processors based on the interconnection-cached network architecture. The optical network operates in a circuit-switching mode. Combined with a packet-switching operation among the circuit-switched optical channels, a high-bandwidth, low-latency network for massively parallel processing results. The design and assembly of a 64-channel experimental prototype is discussed, and operational results are presented.     

Free-space optical cross-connect switch by use of electroholography

An electrically controlled holographic switch is proposed as a building block for a free-space optical interconnection network. The switch is based on the voltage-controlled photorefractive effect in KLTN crystals at the paraelectric phase. It is built of electrically controlled Bragg gratings stored in the volume of the crystal. A compact switch that connects four high-speed fiber-optic communication channels with high efficiency is demonstrated experimentally. The switch performance is investigated and optimized. This switch is extremely attractive for cascaded switching arrays such as those found in multistage interconnect networks.     

Polarization-controlled multistage switch based on polarization-selective computer-generated holograms

We describe a polarization-controlled free-space optical multistage interconnection network based on polarization-selective computer-generated holograms: optical elements that are capable of imposing arbitrary, independent phase functions on horizontally and vertically polarized monochromatic light. We investigate the design of a novel nonblocking space-division photonic switch architecture. The multistage-switch architecture uses a fan-out stage, a single stage of 2 x 2 switching elements, and a fan-in stage. The architecture is compatible with several control strategies that use 1 x 2 and 2 x 2 polarization-controlled switches to route the input light beams. One application of the switch is in a passive optical network in which data is optically transmitted through the switch with a time-of-flight delay but without optical-to-electrical conversions at each stage. We have built and characterized a proof-of-principle 4 x 4 free-space switching network using three cascaded stages of arrayed birefringent computer-generated holographic elements. Data modulated at 20 MHz/channel were transmitted through the network to demonstrate transparent operation.     

All-fiber spot-size transformer for efficient free-space optical interconnecting devices

A new concatenated configuration of optical fibers is proposed to improve misalignment tolerance and working distance in a free-space optical interconnection. The structure is composed of three segments: a thermally expanded core fiber, a coreless silica fiber, and a claddingless graded-index fiber. A numerical analysis for the design of each fiber segment is performed completely, and the device characteristics are experimentally evaluated in terms of the tolerances for a 1-dB coupling-loss increment against longitudinal, transverse, and angular misalignments.     

Diffractive optics applied to free-space optical interconnects

The optimum design of free-space optical interconnection systems utilizing diffractive optics is determined from a practical engineering standpoint for systems ranging from space invariant to fully space variant. System volume is calculated in terms of parameters such as the f-number of the diffractive lens, the wavelength of light, and also the total number, size, and separation of the optical sources and detectors. Performance issues such as interconnection complexity, diffraction efficiency, and signal-tonoise ratio are discussed. Diffractive optics fabricated by electron-beam direct-write techniques are used to provide experimental results for both shuffle-exchange and twin-butterfly free-space optical interconnects.     

Design and implementation of a modulator-based free-space optical backplane for multiprocessor applications

Design and implementation of a free-space optical backplane for multiprocessor applications is presented. The system is designed to interconnect four multiprocessor nodes that communicate by using multiplexed 32-bit packets. Each multiprocessor node is electrically connected to an optoelectronic VLSI chip which implements the hyperplane interconnection architecture. The chips each contain 256 optical transmitters (implemented as dual-rail multiple quantum-well modulators) and 256 optical receivers. A rigid free-space microoptical interconnection system that interconnects the transceiver chips in a 512-channel unidirectional ring is implemented. Full design, implementation, and operational details are provided.     

Optoelectronic neural-network scheduler for packet switches

A novel, to our knowledge, type of packet scheduler that could significantly outperform current state-of-the-art schedulers is presented. The operation and the design of such a scheduler are discussed, and a fully operational experimental implementation is described. The scheduler uses a neural network in a winner-take-all strategy to optimize decisions on the throughput of both a crossbar and a banyan switching fabric. The problems of high interconnection density are solved by use of a free-space optical interconnect that exploits diffractive optical techniques to generate the required interconnection patterns and weights.     

Performance scaling comparison for free-space optical and electrical interconnection approaches

Projected performance metrics of free-space optical and electrical interconnections are estimated and compared in terms of smart-pixel input-output bandwidth density and practical geometric packaging constraints. The results suggest that three-dimensional optical interconnects based on smart pixels provide the highest volume, latency, and power-consumption benefits for applications in which globally interconnected networks are required to implement links across many integrated-circuit chips. It is further shown that interconnection approaches based on macro-optical elements achieve better scaling than those based on micro-optical elements. The scaling limits of micro-optical-based architectures stem from the need for repeaters to overcome diffraction losses in multichip architectures with high bisection bandwidth. The overall results provide guidance in determining whether and how strongly a free-space optical interconnection approach can be applied to a given multiprocessor problem.     

Two-bounce optical arbitrary permutation network

The two-bounce free-space arbitrary interconnection architecture is presented. It results from a series of three-dimensional topological transformations to the Benes network, the minimum rearrangeable nonblocking network. Although functionally equivalent to the Benes network, it requires only two stages of global (spanning multiple chips) optical interconnections. The remaining stages of the modified Benes interconnection network are local and are implemented electronically (on individual chips). The two-bounce network is optimal in the sense that it retains the Benes minimum number of electronic switching resources yet also minimizes the number of optical links needed for global interconnection. Despite the use of higher-order k-shuffle (k > 2) global optical interconnects, the number of 2 x 2 switching elements is identical to the two-shuffle Benes network: there is no need for k x k crossbar switches for local interconnection at each stage. An experimental validation of the two-bounce architecture is presented.     

Locality and decomposition of regular optical interconnection patterns

Methods that a designer can use to optimize the placement of nodes in a large switching network to decrease the requirements on holographic interconnections are investigated. Localized interconnections between subdivided switches are combined with simpler global interconnections. The interconnections between subdivided switches can be implemented by use of metallic traces on smart-pixel arrays. The global interconnections would be provided by optical free-space techniques. Several advantages arise from this procedure: (1) The regular interconnection pattern is decomposed into several pipes (collection of light beams that form a complete pattern) without loss of functionality. (2) The interconnection pattern may be optimized by variation of the placement of the switches in a switching network (e.g., to obtain a minimum deflection angle). (3) The interconnection pattern may be adjusted to the need of an algorithm by an additional parameter (the dimension). The application to photonic switching networks and signal processing is discussed.     

Coded-orthogonal frequency division multiplexing in hybrid optical networks

Future Internet should be able to support a wide range of services containing large amount of multimedia over different network types at a high speed. The future optical networks will therefore be hybrid, composed of different single-mode fibre (SMF), multi-mode fibre (MMF) and free-space optical (FSO) links. In these networks, novel modulation and coding techniques are needed that are capable of dealing with different channel impairments, be it in SMF, MMF or FSO links. The authors propose a coded-modulation scheme suitable for use in hybrid FSO - fibre-optics networks. The proposed scheme is based on polarization-multiplexing and coded - orthogonal frequency division multiplexing (OFDM) with large girth quasi-cyclic low-density parity-check (LDPC) codes as channel codes. The proposed scheme is able to simultaneously deal with atmospheric turbulence, chromatic dispersion and polarisation mode dispersion (PMD). With a proper design for 16-quadrature amplitude modulation (QAM)-based polarisation-multiplexed coded-OFDM, the aggregate data rate of 100 Gb/s can be achieved for OFDM signal bandwidth of only 12.5 GHz, which represents a scheme compatible with 100 Gb/s per wavelength channel transmission and 100 Gb/s Ethernet.     

Design of an optical interconnect for photonic backplane applications

A compact alignment-tolerant interconnect has been developed for use within a prototype modulator-based free-space photonic backplane. The interconnect design encompasses several unique features. Microlens arrays are used, and several beams share each microlens by clustering the optical input-output in a small field about the optical axis of each lens. For simplifying the layout, the optical input and output of each smart-pixel array are clustered separately, thereby allowing a Fourier plane patterned-mirror array to be used in the beam-combination optics. This allows a suitable balance between high interconnection densities and reasonable optical relay distances between adjacent boards to be achieved. The primary advantages of this scheme are the simplicity of the optical design and its alignability, making it ideally suited for high-density interconnection applications.     

256 × 256 Turnover-type free-space multichannel optical switch based on polarization control using liquid-crystal spatial light modulators

Free-space multichannel optical switches using polarization control are attracting interest for future telecommunication networks and interconnection networks in computers. We describe a switching architecture, the turnover type, for such free-space multichannel optical switches. The architecture makes it possible to realize a large-scale and transparent optical switch that is also compact. A 256 × 256 multichannel optical switch based on the architecture is designed and fabricated. To the authors' knowledge, the channel number of the fabricated switch is the largest yet reported among rearrangeable optical switches. Switching operation and signal transmission at 400 Mbits/s are performed successfully with a prototype switch.     

Three-dimensional board-to-board free-space optical interconnects and their application to the prototype multiprocessor system: COSINE-III

A prototype multiprocessor system using three-dimensional board-to-board free-space optical interconnects is constructed for the first time to our knowledge. In the system, 64 processing units form a three-dimensional mesh processor network with the help of bidirectional board-to-board free-space optical interconnects. A theoretical analysis shows that the three-dimensional board-to-board freespace optical interconnects effectively solve common interconnection problems such as wiring congestion, signal delay, and clock skew. The prototype system, COSINE-III, is confirmed to work well as a multiprocessor system. The system is also shown to be easy to extend to a larger and more flexible system.     

Free-space optical crossbar network integrated in a single block of LiNbO3 crystal

A free-space optical crossbar network integrated in a single block of LiNbO(3) crystal is proposed, which consists of stages of 2×2 switches making use of the electro-optic effect of crystal and in-between routing devices for permutation based on double refraction and internal double reflection on interfaces. Two basic configurations are suggested. A control algorithm for the crossbar network is discussed, which may control a nonblocking interconnection between any input and output. The integrated crossbar network is low energy loss, nonblocking, easy to assemble, and insensitive to environment. A 3×3 crossbar network is designed and the experiment is demonstrated.     

Analytical models of blocking probability for multi-granularity cross-connect-based optical networks

Multi-granularity optical cross-connect (MG-OXC)-based optical network is a promising optical network architecture as it is capable of flexible switching at different granularity levels. In MG-OXC-based optical networks, wavelength conversion (WC) capability and the number of usable add/drop ports of the nodes are two key factors affecting its performance. Two analytical models of blocking probability for MG-OXC-based optical networks both without WC capability and with sparse WC capability are proposed, exploiting Erlang?s loss formula and birth?death process. Based on the models and simulation, the impact of WC capability and the number of add/drop ports on the blocking probability are investigated. Three kinds of granularities (i.e. fibre, waveband and wavelength) are considered in MG-OXC nodes to reduce the complexity and size of switch fabric. Both the analytical and simulation results are given on two network topologies under dynamic traffic patterns. Simulation results show that the proposed models are accurate and effective for the analysis of blocking probability in MG-OXC-based optical networks.     

Sliding-banyan network performance analysis

The sliding-banyan (SB) network employs an interleaved multistage shuffle-exchange topology, implemented with a three-dimensional free-space interconnection architecture that connects a multichip backplane to itself. Surface-normal emitters and detectors, which compose the stages' input-output, are spatially multiplexed within the same chip location, along with electronic control and switching resources. A simple deflection self-routing scheme minimizes internal contention, providing efficient use of switching and interconnection resources. The blocking performance of the SB is quantified through simulations based on realistic nonuniform traffic patterns. Results show that the SB architecture requires significantly fewer resources than other self-routing banyan-based networks. The multistage-switching and interconnection-resource requirements are close to the theoretical minimum for nonblocking networks, and the SB's distributed self-routing control resources grow only approximately linearly with the number of nodes, providing good scalability.